INTRODUCTION
Smartphones have transformed modern life by enhancing convenience and efficiency in information retrieval, social networking, education, and leisure activities. For children and adolescents, smartphones serve as educational tools and information resources and help expand their social relationships. However, despite these advantages, studies have linked excessive smartphone use to adverse effects on children’s psychological and behavioral development [1,2]. Research also suggests children’s vulnerabilities and characteristics contribute to their dependence on smart devices. Scholars have identified internalizing and externalizing behavioral problems, including depression, anxiety, attention deficits, and self-regulation issues, as precursors to smartphone dependence [3,4]. This relationship appears to be bidirectional, as smartphone dependence can also exacerbate these behavioral issues, leading to a mutually reinforcing cycle.
The developmental model of addiction [3,5] and the co-occurrence hypothesis [6] suggest that children and adolescents with internalizing problems may use smartphones to escape negative emotions. In the case of externalizing behavior problems, the relationship with smartphone dependence may be more complex, as these problems can both result from and contribute to excessive smartphone use [7-9]. Children with externalizing issues, such as heightened sensitivity to rewards, impulsivity, and inhibitory control difficulties, are at risk of smartphone dependence [3,10]. In turn, excessive smartphone use can intensify impulsivity and reward sensitivity, further aggravating externalizing behavior problems [11].
While longitudinal studies provide some evidence for this bidirectional relationship, they remain limited. For instance, Poulain et al. [7] found that internalizing and externalizing problems predicted increased smartphone addiction symptoms over one year. However, they also reported that increased smartphone addiction predicted a rise in externalizing but not internalizing behavior problems. However, Poulain et al.’s study [7] did not account for the autoregressive effects of smartphone dependence, which are essential to understanding how prior levels of dependence predict future levels.
The autoregressive cross-lagged model (ACLM) is a robust statistical method for analyzing bidirectional relationships over time. It accounts for autoregressive effects (i.e., the stability of a variable over time) and cross-lagged effects (i.e., the influence of one variable on another over time). Previous studies have used ACLM to examine smartphone dependence with school adjustment, depression, and peer relationships among children and adolescents [5,12,13]. However, no study has applied ACLM to explore the longitudinal relationship between smartphone dependence and externalizing behavior problems.
Given the potential for smartphone dependence and externalizing behavior problems to mutually influence one another over time, this study employs ACLM to examine their longitudinal and bidirectional association. The study seeks to clarify the interaction and evolution of smartphone dependence and externalizing behavior problems in children by controlling for gender, age, and co-occurring internalizing problems. The findings aim to inform interventions targeting behavioral issues associated with smartphone use in children.
METHODS
Participants and procedure
The study recruited children aged 7-12 from South Korea for a longitudinal assessment conducted across four waves at six-month intervals, spanning from June 2021 and December 2022. The analysis included data from 338 of the 379 children who participated in at least two waves. Among these, 169 children completed all four waves, 86 participated in three, and 83 completed two. The institutional review board of the Dongduk Women’s University approved the study (DDWU2103-03), ensuring ethical compliance. The children received the study questionnaires at school, which they delivered to their mothers after obtaining prior consent to participate. Mothers completed the questionnaires at home, sealed them, and returned them via their children to the school, where the researchers collected them. As a token of appreciation after data collection at each time point, researchers gave mothers brief feedback on their child’s behavior problems based on the Korean Version of the Child Behavior Checklist (K-CBCL) results. Participation was entirely voluntary, with no penalties for withdrawal at any point.
The final sample included 178 girls and 160 boys, with an average age of 9.15 years (standard deviation=1.63).
Measures
Child Smartphone Overdependence Observer Scale
The study assessed smartphone dependence in children using the Child Smartphone Overdependence Observer Scale, developed by the Korean Information Society Agency [14]. The scale consists of nine items across three factors: self-control failure, salience, and serious consequences. Respondents rated each item on a four-point Likert scale, with higher scores indicating greater smartphone dependence. At the time of the scale’s development, Cronbach’s α for the total score was 0.75, and in the current study, Cronbach’s α for the total scale was 0.93, indicating good internal consistency.
Korean Version of the Child Behavior Checklist for Ages 6-18
The K-CBCL 6-18 [15] is a parent-reported questionnaire assessing the behavioral and emotional problems of school-aged children. Achenbach and Rescorla [16] initially developed the checklist, which researchers later translated and standardized for Korean populations [15]. The K-CBCL is a widely used measure for assessing psychopathology in children aged 6 to 18 years. It includes 20 items on social competence and 113 items that assess behavioral problems, rated by parents on a three-point scale. In the Korean standardization study, Cronbach’s alpha for each scale ranged from 0.62 to 0.84. The present study focused on the internalizing and externalizing subscales of the K-CBCL. The internalizing subscale reflects behaviors such as anxiety and withdrawal, while the externalizing subscale captures aggression and rule-breaking
Data analysis
We used SPSS 26.0 (IBM Corp.) for descriptive statistics, baseline internalizing behavior problems, and correlations between smartphone dependence and externalizing behavior problems across time. Repeated measures of analysis of variance (ANOVA) assessed changes in variables over time. We evaluated missing data (1%-10% at the item level) using Little’s Missing Completely at Random test, which confirmed that data were indeed missing completely at random (χ2 [170]=188.389, p=0.159). We imputed missing values using the expectation-maximization procedure in SPSS [2].
To examine the longitudinal relationship between smartphone dependence and externalizing behavior problems, we conducted ACLM using SPSS AMOS and included gender, age, and internalizing behavior problems at time one (T1) as covariates. Analyses employed the full information maximum likelihood method to handle missing data. We assessed the model fit using chi-square (χ2) statistics, root mean square error of approximation (RMSEA) (with 90% confidence intervals [CIs]), comparative fit index (CFI), and the Tucker-Lewis Index (TLI), with fit criteria set at CFI/TLI≥0.90 and RMSEA≤0.08 [17].
We compared an unconstrained cross-lagged model (Model A) with four alternative models: 1) Model B, in which we constrained all cross-lagged paths between smartphone dependence and externalizing behavior problems to 0, 2) Model C, in which we set the paths from prior smartphone dependence to later externalizing behavior problems to 0, 3) Model D, in which we set the paths from the prior externalizing behavior problem to the later smartphone dependence to 0, and 4) Model E, in which we constrained cross-lagged paths between the two variables to be equal.
RESULTS
Preliminary analysis
Descriptive statistics and correlations for smartphone dependence and externalizing behavior problems across time points are in Table 1. All variables met normal distribution assumptions, with skewness and kurtosis values below the thresholds of 2 and 4 [18]. Correlations between smartphone dependence across time points ranged from 0.70 to 0.87, while those for externalizing behavior problems ranged from 0.70 to 0.82. Concurrent and longitudinal correlations between smartphone dependence and externalizing behavior problems ranged from 0.22 to 0.43.
Repeated measures ANOVAs revealed significant changes over time: smartphone dependence increased significantly, F(3, 335)=13.98, p<0.001, while externalizing behavior problems decreased significantly, F(3, 335)=10.84, p<0.001. Post hoc analyses showed that smartphone dependence increased significantly between the first and second periods while externalizing behavior problems decreased significantly from the second to third periods.
Autoregressive cross-lagged model
Model fit indices are in Table 2. The unconstrained model (Model A) demonstrated adequate fit: χ²=110.628, df=30, p<0.001; RMSEA=0.084 (90% CI: 0.068-0.101), p-close=0.000; CFI=0.941; TLI=0.871. Chi-square difference tests indicated that Model A was significantly superior to Models B and C (χ2=32.33, df=6, p<0.001; χ2=27.16, df=3, p<0.001). Although Models D and E performed slightly better in RMSEA, we found no significant differences in CFI or TLI, so we selected Model A as the final model.
The results (Figure 1) showed strong stability over time for smartphone dependence (β=0.75-0.86, p<0.001) and externalizing behavior problems (β=0.67-0.78, p<0.001). Cross-lagged analyses revealed that earlier smartphone dependence significantly predicted subsequent externalizing behavior problems (T1 → T2: β=0.19, p<0.001; T2 → T3: β=0.12, p<0.001; T3 → T4: β=0.10, p=0.005). However, earlier externalizing behavior problems did not significantly predict subsequent smartphone dependence.
Covariates were also significant: we found a negative association between gender and smartphone dependence (β=-0.12, p=0.020) and externalizing behavior problems (β=-0.12, p=0.005). Age had a positive association with smartphone dependence (β=0.22, p<0.001). Internalizing behavior problems at T1 were significantly related to smartphone dependence (β=0.26, p<0.001) and externalizing behavior problems (β=0.63, p<0.001).
DISCUSSION
This study explored the longitudinal relationship between smartphone dependence and externalizing behavior problems in children. We found that smartphone dependence increased over time while externalizing behavior problems decreased, consistent with previous longitudinal research that reported similar trends [7,11]. Notably, we observed these changes during the COVID-19 pandemic, a period linked to increased smartphone use among children [19].
The results of the autoregressive cross-lagged interaction between children’s smartphone dependence and externalizing behavior problems are as follows. First, smartphone dependence and externalizing behavior problems exhibited strong autoregressive effects, reflecting high temporal stability. Second, cross-lagged analyses revealed that smartphone dependence significantly predicted later externalizing behavior problems, while externalizing problems did not predict smartphone dependence. This finding suggests that smartphone dependence may be a causal factor in developing externalizing behaviors rather than these behaviors mutually reinforcing one another.
Scholars have long debated the complex interplay between smartphone dependence and behavioral problems in children, with efforts to discern causal directions. For internalizing problems, theories such as the Reward Model of Addiction [4] and the Developmental Model of Addiction [3] suggest that smartphone dependence may serve as a coping mechanism to avoid negative emotions [5]. In contrast, the comorbidity hypothesis [6] can better explain externalizing problems. It posits that smartphone dependence shares a common vulnerability with externalizing behaviors [10] or exacerbates irritability and rule-breaking behaviors when overdependence is restricted [20,21]. Recent longitudinal studies support the notion that smartphone dependence is a significant driver of externalizing behaviors. For example, studies have shown that smartphone overdependence predicts aggression and rule-breaking behaviors in children and adolescents [7,22,23].
The current study adds to this body of literature, demonstrating that earlier smartphone dependence significantly predicts subsequent externalizing problems across multiple time points. The mechanism underlying this relationship is likely multifaceted. Smartphone dependence may impede the development of emotional and behavioral self-regulation, driven by the instant gratification facilitated by digital interactions [24,25]. Excessive use of smartphones may also negatively impact parent-child interactions, leading to harsher parenting practices or reduced parental supervision, known risk factors for externalizing behaviors [23,26]. Additionally, exposure to violent or aggressive content on smartphones may contribute to behavioral modeling effects, increasing the likelihood of aggressive or rule-breaking behaviors [27]. We need further research to identify the specific pathways through which smartphone dependence leads to externalizing behavioral problems.
However, this study has several limitations. First, the findings may lack generalizability due to the relatively small sample size, reliance on parent-reported measures, and potential selection bias from families who completed multiple assessments. In particular, further research incorporating data from multiple sources—such as child self-reports, teacher assessment, and objective behavioral observations—would enhance the validity of the results. Second, this study focused on externalizing behavior problems. However, given the potential interaction between internalizing and externalizing behavior problems, an integrated exploration of both types would offer a more comprehensive understanding of how smartphone dependence and these behaviors interact over time. Third, we did not examine how different types of smartphone use (e.g., gaming, social media, educational apps) might affect externalizing behavior problems in distinct ways. Analyzing behavioral problems by smartphone activity type could yield insights for more targeted interventions. Finally, future research should investigate gender differences in the relationship between smartphone dependence and behavioral problems, as suggested by previous studies employing network analysis [21].
In conclusion, despite its limitations, this study is noteworthy for employing a repeated-measures ACLM to establish a causal link between childhood smartphone dependence and externalizing behavior problems. The findings suggest that smartphone dependence is not merely a habitual behavior but a contributing factor to subsequent externalizing problems. These results emphasize the importance of parents and educators actively managing children’s smartphone use through evidence-based strategies [28]. For instance, structured screen time guidelines can help children balance digital and offline activities, while parental education programs can equip caregivers with skills to monitor and guide appropriate usage. Additionally, school-based initiatives, such as digital literacy programs and counseling services, can further support efforts to mitigate the negative behavioral impacts of smartphone dependence. At a broader level, policy initiatives should aim to foster healthy digital device habits in children by implementing screen time guidelines and designing targeted educational programs.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Print
